You are on page 1of 14

International Journal of Computer Science and

Engineering (IJCSE);
ISSN Print: 2278-9960;ISSN Online:2278-9979;
Vol. 8, Issue 3, Apr - May 2019; 15-28
© IASET

PREDICTION OF PERSONALITY THROUGH SIGNATURE AND HAND GEOMETRY

Vaishali R. Lokhande1, Santwana G. Mishra2 & Bharti W. Gawali3


1,3
Research Scholar, Department of Computer Science and I.T, Dr. Babasaheb Ambedkar Marathwada University,
Aurangabad, Maharashtra, India
2
Research Scholar, Department of Education, Dr. Babasaheb Ambedkar Marathwada University, Aurangabad,
Maharashtra, India
ABSTRACT

Human hand and signature both endow various personalities of an individual that depicts the distinctiveness of an
individual. The present work recognized different characteristics through features of hand namely palm shape ratio and
finger index proportional ratio for categorization of hand shape as Air, Fire, Water, and Earth which also involved the
features of signature such as curved start, ending stroke, underscore, the appearance of the dot on letter and streaks
disconnected. The fusion of these two modalities with respect to predictive models and Big Five Factors; produces an
88.0% accuracy rate and is found to be more efficient in predicting the personality of an individual.

Big Five Factors which encompasses personality related traits and which include both positive and negative
characteristics of an individual. The various factors present in Big Five were used for the evaluation; performed on the
result of traits of signature and palm through matching and replacing factor and algorithmic steps. This research work
focuses on the predictive models which define various personalities of 25 subjects.

KEYWORDS: Big-Five Factor Model, Feature Set, Personality Prediction, Predictive Model

Article History
Received: 23 Apr 2019 | Revised: 02 May 2019 | Accepted: 08 May 2019

INTRODUCTION

The personality focuses on behavioral or physiological characteristics of the person. Now a day’s personality
identification is very popular in every field and it can be characterized through various modes such as handwriting, speech,
face, hand geometry, signature, walking style, big five personality traits and many more. The present research depicts the
work done using bimodal personality prediction system (bpps).

Every person has its own signature which renders the personality traits. The signature analysis is a part of
graphology theory; graphologists testify the various qualities which includes attitudes, personality, and sentiments. These
aspects were described with the help of appearance of dots on the letter as in i, j; ascending bottom line; single or double
underline; ending stroke; starting stroke; the appearance of letter t; streaks; shapes of each letter and so on. Image
processing drives the method of extracting features and recognizes the pattern. In image processing the whole signature is
judged as an image so that detection can be achieved through the numerous stages of conversion of images; such as
arithmetical vector into binary, thresholding of the value and so on; which perk up the superiority of the image. This also
emulates through feature extraction and pattern recognition.

www.iaset.us editor@iaset.us
16 Vaishali R. Lokhande, Santwana G. Mishra & Bharti W. Gawali

Human hand depicts with different temperaments that rule the phenomenon and also portrays the personality of
the individual. In the ancient age, the astrologists were predicting the personality through hand shape, hand length and
width called Chirognomy; and also through palm lines called Chiromancy in the language of palmistry. These categories
diagnose hidden health problems, hidden skill; foretell fortune, best profession, reveals character traits and many more.
Handshapes are classified into Fire, Air, Water, and Earth hand; these are also pronounced as elemental or temperamental
hand shapes. These are believed to signify character traits corresponding to the type defined.

Artificial Neural Network

The central nervous system of human brain theme has been utilized in Artificial Neural Network (ANN)
sometimes called as biological neural networks. The nervous systems are capable of machine learning as well as pattern
recognition. The system of interconnected neurons is one of the main aspects of the ANN. To describe the attributes of
computing, the artificial neural networks go by many names such as connectionist models, parallel distributed processors,
or self-organizing system [1]. The name Neural is connected to neurons, which is an essential part of the biological nervous
system. Human nervous system developed the information as it receives from nerves. Every neural network encompasses
the interconnected neurons which are configured for a specific application.

The advantage is that the neural system can learn new input-output patterns and adjust the system parameters.
With such features, an artificial neural system has great potential in performing applications such as speech, image
recognition and prediction of various characteristics where intense computation can be done in parallel and the
computational elements are connected by weighted links.

Work was done so far

Most of the researchers determined the personality through handwriting only.

[1] [2] described a method for personality identification of a person via baseline of the sentence, pen pressure, the
representation of the letter ‘t’ and letter ‘y’ present in an individual’s handwriting. Artificial Neural Network was applied
for prediction of personality traits. Similarly [3] considered baseline, slant letter, pen pressure and the letters ‘i’ and ‘f’ is
identified. Various features of handwriting were considered [4] for prediction of human personalities such as words and
letters bluntness, size of the letters, baseline, pen pressure, the spacing between letters and words using support vector
machine.

It was observed from the literature survey that, maximum work based on handwriting, facial expression, signature,
palm, speech and many more, for prediction of personality. Very few researchers considered the fusion work on
handwriting and signature for personality identification. [5] portrays the personality prediction of the individual through
feature extraction methods which include statistical, structural and global transformation; theobtained values were
classified using artificial neural network and got 90% accuracy for the prediction of personality traits.

A fusion of digit of each character and signature were implemented [6] using multi-structure algorithms and
artificial neural networks. This research paper considered the characters present in the application form and the signature
image based on nine features; extreme margin, dot structure, separate, streaks disconnected, curve starts, end streaks,
middle streaks, underline and shell with good accuracy rate. [7] Completed the study on the handwriting and signature
recognition using multiple artificial neural network and multi-structure algorithm and also done a comparative analysis of
various techniques.

Impact Factor (JCC): 6.0127 NAS Rating: 3.80


Prediction of Personality Through Signature and Hand Geometry 17

Automated medical palm systems were implemented to diagnose different diseases present in the human body
through palm lines and fingers. 92 images were correctly predicted from 100 and result so obtained with an accuracy rate
of 94% [8], [9]. To recognize human behavior, human nature, human personality, health, emotions, and other features; the
well -known field called palmistry were applied by [10].

The prediction of an individual’s first impression through visual and audio features accompanied by individuals
Big-Five Personality Traits namely Openness, Conscientiousness, Extroversion, Agreeableness and Neuroticism
(OCEAN). The work had been done using Deep Bimodal Regression LSTM model that extracts temporally ordered visual
and audio features from a video clip and outputs five scores with a range from 0 to 1. The model was trained and evaluated
on HD Youtube videos provided by ChaLearn LAP APA2016 dataset; and achieved excellent performance accuracy rate
with 90-91% [11]. [12] presented a technique on CNN model interpretability, which had been combined with face
detection and Actions Unit (AUs) recognition systems. This system focused only on video frames, by discarding the audio
information and then performed a deep study on the interpretability of CNN models trained exclusively from video frames
that revealed the facial feature detectors namely, eyes, eyebrows, nose and mouth from the intermediate layers of the CNN
models to perform apparent personality trait regression. Thus achieved good results and demonstrate that face endow with
discriminative information for personality trait presumption and the interior CNN depiction analyzed key face section like
eyes, mouth, and nose. The development of an audiovisual deep residual network for the multimodal apparent personality
trait of people in an end-to-end manner had been implemented by [13]. This approach obtained very high performance for
all traits through OCEAN and identified relevant features for the task of predicting apparent personality traits. The
resultant won third place in the ChaLearn First Impression Challenge with a test accuracy of 0.9109.

Experimental Work

The flow for experimentation is as follows:

Image Acquisition and Database Creation

The signature and hand images were taken from the 25 subjects belonging to different age groups at the same
time. Firstly, hand images were captured from the camera. These captured images were stored in a.jpeg file format. After
capturing hand images, the subjects were requested to sign on a white paper in the specified rectangular area. The images
were scanned at 600*600 dpi resolution and resized to 200*100 long. Thus the databases were created from 25 subjects for
50 hand images (i.e. 25 images of left and right hand) and 6 samples of signatures. Thus total of 300 samples of signatures
were used for the further processing.

Preprocessing Stages of Hand Images and Signature

There are various types of images and these images need to be gone through different processing steps.
Preprocessing consists of [14]:

• RGB image: This image is also called a true color image. It is stored in m by n, 3 data array that defined as Red,
Green and Blue color components for every pixel. When all the intensities values of RGB are equal then it
produces black (0, 0, 0) or white (1, 1, 1) color.

• Grayscale image: This is a data matrix whose value represents intensities within some range. The MATLAB
function rgb2gray( ) was applied to convert image from color image to gray image.

www.iaset.us editor@iaset.us
18 Vaishali R. Lokhande, Santwana G. Mishra & Bharti W. Gawali

• Binary image: This image contains only two pixels i.e. black and white and abbreviated as BW image. The gray
image can be converted to a binary image using im2bw( ) function in MATLAB.

Morphological Operations on Image

There are various morphological functions are present in MATLAB such as contrast enhancement, noise removal,
thinning, skeletonization, filling, segmentation, dilation, erosion and so on. Among them we had applied; imfill( ), imopen(
) and imerode( ) on the scanned signature and hand images.

Segmentation

• Both the images were segmented separately; the signature image is segmented into two; Vertical and Horizontal
segmentation, which includes left-right and upper-middle- top segments respectively.

• In a similar manner, the hand image is also divided into sections; palm and fingers, which includes length-width-
ratio and length-ratio respectively. These segmented images were helpful for feature extraction.

Feature Extraction

Feature extraction is used in image processing and pattern recognition to obtain the most pertinent information
from the distinctive data and accumulate the data to reduce dimensionality.

• The essential groups of the feature were extracted from both the images, after performing morphological
operations. To predict personality, total of five features were considered for signature. Table I elaborates five
features.

• Hand also plays a vital role in the personality prediction, thus the evaluation had been done through;

Palm measurements which include palm length (pl), palm breadth (pb), and palm shape ratio
(pb/pl)……………………………………………… ….(1)

The finger palm ratio which include finger length (fl)/ palm breadth (pb) and finger length (fl)/ palm length (pl).
To enhance the evaluation finger index proportional ratio were calculated using the formula:finger length/ (palm
breadth+palm length)/2 ………… …….(2)

Palm features were calculated through the obtained values of palm shape ratio and finger index proportional ratio
using formula (1) and (2). After applying and evaluating these two formulas and another corresponding formula, the
resultant values were helpful for the categorization of elemental hand shapes such as earth hand, water hand, air hand, and
fire hand.

Both signature and hand express the personalities of an individual based on the feature set and are shown in Table
1 and Table 2. The feature of signature of each subject which has been illustrated through binary values 0 and 1 were
passed through Artificial Neural Network for predicting the personality with respect to the type of hand category (using
formula (1) and (2)) is explained in Table 3.

Impact Factor (JCC): 6.0127 NAS Rating: 3.80


Prediction of Personality Through Signature and Hand Geometry 19

Table 1: Feature Set of Signature and Its Personality

Table 2: Feature Set of Palm and Its Personality

www.iaset.us editor@iaset.us
20 Vaishali R. Lokhande, Santwana G. Mishra & Bharti W. Gawali

Table 3: Feature Set of Signature of each Subject and their Category of Hand Shape using Corresponding
Formulae

Signature, Palm and Big Five Factor Model

Scientific psychology has developed a high-level vision of personality encompassing traits, sets of stable
dispositions towards action, belief and attitude formation. Big Five Personality traits is a well known and one of the
influential example of this approach also called as Five Factor Model as they encompass a large portion of personality-
related terms which includes; Openness, Conscientiousness, Extroversion, Agreeableness, and Neuroticism.

These five factors are not necessarily traits in and of themselves, but factors in which many related traits and
characteristics fit. For example, the factor agreeableness includes terms like generosity, friendliness, and warmth (on the
positive side); and assertiveness and temper (on the negative side). All of these traits and characteristics make the broader
factor of “agreeableness”. In a similar manner, all Five Factors describes its own positive face and negative face.

Table 3 is sufficient to categorize the personality of an individual; but for getting a better result with respect to
Signature and Palm; the mapping is done through Big Five Factors thus, Predictive Model-1 and Predictive Model-2 is
implemented through algorithmic steps.

The Table 4 and 5 describe distinct ranges of Signature and Palm which has been obtained from the mean value of
each feature set i.e Curved start, Ending stroke, Ascending bottom, Streaks disconnected, Appearance of dot for signature
and Palm ratio, Finger Index for hand respectively; of each 25 subjects. An AND operations were performed through
ranges of each feature set which helps in obtaining the sub-categories of Big-Five Factor such as Preserver, Moderate,
Explorer in Openness Category; Flexible, Balanced, Focused in Conscientiousness Category; Introvert, Ambivert,
Extrovert in Extroversion Category; Challenger, Negotiator, Adapter in Agreeableness Category and Resilient, Responsive,
Reactive in Neuroticism Category.

Impact Factor (JCC): 6.0127 NAS Rating: 3.80


Prediction of Personality Through Signature and Hand Geometry 21

Table 4: Ranges of Feature Set of Signature and Its Result Based on AND Operation

Table 5: Ranges of Feature Set of Palm and its Result based on AND operation

• The ranges of Signature those are in between 0.08 – 0.12 of Curved start AND 0.03 – 0.08 of Ending stroke AND
0.02 - 0.06 of Ascending bottom line AND 0.13 – 0.18 of Streaks disconnected AND 0.01 – 0.05 of Appearance
of dots. Similarly, ranges of Palm those are in between 0.18 - 0.92 of Palm ratio AND 0.36 – 0.80 of Finger Index
then categorization of five-factor model is:

• In Openness the subject is Preserver

• In Conscientiousness the subject is Flexible

• In Extroversion the subject is Introvert

• In Agreeableness the subject is Challenger

• In Neuroticism the subject is Resilient.

• The ranges of Signature those are in between 0.12 – 0.16 of Curved start AND 0.08 – 0.12 of Ending stroke AND
0.06 - 0.10 of Ascending bottom line AND 0.18 – 0.23 of Streaks disconnected AND 0.05 – 0.09 of Appearance
of dots. Similarly, ranges of Palm those are in between 0.92 - 1.10 of Palm ratio AND 0.80 – 1.16 of Finger Index
then categorization of five-factor model is:

• In Openness the subject is Moderate

• In Conscientiousness the subject is Balanced

• In Extroversion the subject is Ambivert

www.iaset.us editor@iaset.us
22 Vaishali R. Lokhande, Santwana G. Mishra & Bharti W. Gawali

• In Agreeableness the subject is Negotiator

• In Neuroticism the subject is Responsive.

• The ranges of Signature those are in between 0.16 – 0.19 of Curved start, 0.12 – 0.17 of Ending stroke, 0.10 - 0.15
of Ascending bottom line, 0.23 – 0.29 of Streaks disconnected and 0.09 – 0.13 of Appearance of dots. Similarly,
ranges of Palm those are in between 1.10 - 1.28 of Palm ration AND 1.16 – 1.51 then categorization of five-factor
model is:

• In Openness the subject is Explorer

• In Conscientiousness the subject is Focused

• In Extroversion the subject is Extrovert

• In Agreeableness the subject is Adapter

• In Neuroticism the subject is Reactive.

• Table 6 below describes the personality traits of the Big Five Factor of Signature and Palm. The comparative
study has done through obtained ranges as shown in Table 4 and 5 above to enhance the personality concept
through the Big Five Factor. The match factor has calculated and obtained the overall accuracy of 71.2%. The
present research considered this as Predictive model-1.

Impact Factor (JCC): 6.0127 NAS Rating: 3.80


Prediction of Personality Through Signature and Hand Geometry 23

Table 6: Evaluation of Big Five Factor Traits through Signature and Palm

To improve the overall accuracy rate, Predictive Model-2 is implemented where mapping has been done through
Predictive Model-1 and shown diagrammatically in Figure 1 below:

www.iaset.us editor@iaset.us
24 Vaishali R. Lokhande, Santwana G. Mishra & Bharti W. Gawali

Figure 1: Mapping between Predictive Model-1 and Predictive Model-2

Predictive Model-2 is prepared through the results of Predictive Model-1. Through obtained ranges (Table 4 and
5), the categorization of Mid-values is produced which is helpful for matching and replacing of the values and is shown in
Table 7 and 8 and for signature and palm respectively.

Table 7: Categorization of Mid-Values for Mapping with Big Five Factor through Signature

Impact Factor (JCC): 6.0127 NAS Rating: 3.80


Prediction of Personality Through Signature and Hand Geometry 25

Table 8: Categorization of Mid-Values for Mapping with Big Five Factor Through Palm

Algorithmic steps-I were applied which describes the matching and replacing steps.

Algorithmic Steps for Matching and Replacing

Step 1: Each subject’s highlighted sub-category of Big Five Factor has compared with equivalent non-
highlighted sub-category by observing their mid-values as described in Table 7 and 8.

Step 2: These mid-values were again added to obtain the resultant value.

Step 3: This resultant value is perfectly matched or not matched was observed through the ranges of predefined
values.

Step 4: If it is matched then replaced by resultant value otherwise remain the same.

Thus, an overall accuracy rate is calculated by performing AND-Operations using Table 7 and 8 on big five-factor
of predictive model-1; the match factor was evaluated and accordingly accuracy rate was obtained which is helpful to
achieve the result of 88.0% as described in Table 9.

www.iaset.us editor@iaset.us
26 Vaishali R. Lokhande, Santwana G. Mishra & Bharti W. Gawali

Table 9: Evaluation of Big Five Factor Traits through Signature and Palm

Result Analysis

The study involved the analysis of the personality of the selected data set consisting of 300 samples of signature
and hand images that were drawn from 25 peoples belonging to various age groups. The images were captured for both left
and right hands. Thus a total of 50 palm images were analyzed including 25 left palm images and 25 images of the right
palm. The analysis also included 6 samples of signatures for each selected subject. These datasets were used for predicting
personality traits. Various functions of MATLAB were used for preprocessing the dataset. The researchers also conducted
the Big Five personality test to be used for validation of the final result.

The final mapping of the dataset was annotated with the Big Five Personality Traits. The researchers implemented
two predictive models through performing AND-operations on the ranges and obtained the overall accuracy rate of 71.2%
for Predictive Model-1 and 88.0% for Predictive Model-2.

The results of Predictive Model-1 are helpful for getting improved accuracy rates. The mapping has done through
the match factor between each trait of Big-Five Factor with each trait of Predictive Model-1. Hence, obtained good
accuracy rate with several common personalities for the Predictive Model-2.

CONCLUSIONS

Successful prediction of personality is one of the challenges faced by the modern world. Along with the age-old

Impact Factor (JCC): 6.0127 NAS Rating: 3.80


Prediction of Personality Through Signature and Hand Geometry 27

systems, newer methods and models are being developed to successfully predict the personality. In the present study, the
researchers have developed a predictive model for predicting the common and prominent personality traits of people by
analyzing the pictures of palm and signatures which is found to be efficient and reliable. This is done through:

• Categorization of Hand-shape into Earth, Air, Water and Fire using palm shape ratio and finger index proportional
ratio. And,

• Features of Signature namely Curved Start, Ending Stroke, Underscore and Appearance of Dot on letter and
Streaks Disconnected.

REFERENCES

1. H.N Champa, K.R Ananda Kumar,” Artificial Neural Network for Human Behavior Prediction through
Handwriting Analysis"; International Journal of Computer Applications (0975 8887) Volume 2 No.2, May 2010.

2. Champa H N, K R Ananda Kumar,” Automated Human Behavior Prediction through Handwriting Analysis";
2010 First International Conference on Integrated Intelligent Computing.

3. P. K. Grewal, D Prashar;”Behavior Prediction Through Handwriting Analysis"; IJCST Vol. 3, Issue 2, April -
June 2012.

4. S. Prasad, V.K. Singh, A. Sapre,”Handwriting Analysis based on Segmentation Method for Prediction of Human
Personality using Support Vector Machine"; International Journal of Computer Applications (0975 8887) Volume
8 No.12, October 2010.

5. Prof. Seema Kedar, Ms. Vaishnavi Nair, Ms. Shweta Kulkarni,”‘Personality Identification through Handwriting
Analysis: A Review”, International Journal of Advanced Research in Computer Science and Software
Engineering (IJARCSSE), Vol-5, Issue-1, pp No: 548- 556, January2015.

6. Esmeralda C. Djamal, Sheldy Nur Ramdlan, Jeri Saputra, “Recognition of Handwritten Based on Signature and
Digit of Character using Multiple of Artificial Neural Networks in Personality Identification”, Information System
International Conference (ISICO), 2-4

7. December 2013.

8. Pooja Ravishankar Tarare, Dipti Jadhav, “Study of Personality Prediction Based on Handwriting and Signature
Recognition using Multiple Artificial Neural Network And Multi-Structure Algorithm”, International Journal of
Pure And Applied Research in Engineering And Technology (IJPRET), Volume 3(9), pp No. 681-693, 2015.

9. Shivali Soni, Dr. Kapil Gupta, “A Study of Automation in Palmistry”, International Journal of Advanced
Research in Computer Science and Software Engineering (IJARCSSE), Vol-5, Issue-5, pp No: 313-317, May
2015.

10. Disha Desai, Mugdha Parekh, Devanshi Shah, Prof. Vinaya Sawant, Prof. Anuja Nagare, “Automated Medical
Palmistry System based on Image Processing Techniques”, International Journal of Advanced Research in
Computer Science and Software Engineering (IJARCSSE),Vol-5, Issue-1, pp No: 725-729, January 2015,

www.iaset.us editor@iaset.us
28 Vaishali R. Lokhande, Santwana G. Mishra & Bharti W. Gawali

11. Dr. Kanchana Kohombange, “Recognizing Human Nature and Behaviors through Hand Analysis”, International
Journal of Interdisciplinary Research and Innovations, Vol. 3, Issue 4, pp: (10-16), pp No.10-16, October -
December 2015.

12. Karen Yang, Noa Glaser, “Prediction of Personality First Impressions With Deep Bimodal LSTM”, pp No. 1-10.

13. Carles Ventura, David Masip, Agata Lapedriza, “Interpreting CNN Models for Apparent Personality Trait
Regression”, CVPR workshop Proceeding paper, An Open Access version provided by the Computer Vision
Foundation, pp No.55-63.

14. Yagmur Guichiturk, Umut Guiclii, Marcel A.J. van Garven, Rob van Lier, “Deep Impression: Audiovisual Deep
Residual Networks for Multimodal Apparent Personality Trait Recognition”, 16 Sept 2016.

15. Karbhari Kale, Ramesh Manza, Vikas Humbe, Pravin Yannawar, Ganesh Manza, “Understanding MATLAB”,
SPD Publications.

16. Kyriaki Kalimeri, Bruno Lepri, Fabio Pianesi, “Going Beyond Traits: Multimodal Classification of Personality
States In The Wild”, ICMI’13, Sydney, Australia, December 9-13, 2013.

17. http://www.handresearch.com/hand-sign-tutor/hand-shapes/elemental-hand-shapes.htm

18. P.J. Howard, P.L. Medina and J.M. Howard, “The Big Five Locator: A Quick Assessment Tool for Consultants
and Trainers”, from The Annual, Developing Human Resources, by J.William Pfeiffer and David Leonard
Goodstein, Vol. 1,, “Training”, 1996, pp. 119-22.

Impact Factor (JCC): 6.0127 NAS Rating: 3.80